We have requirement of seperate database for each child portal. So in part of solution we have seperate database having core dnn table in all of them. further the sql provider is modified to provide a different connection string based on portalID

Now we have one code and n number of client database with one application cache. 

The problem is with cache items. The cache key may be same for different client( hypothetically same tabID). This may lead to incorrect behaviour in terms of data in the cache depending  on client the request comes first. The solution we are implemanting  is to have append the uniquename + portalID in cache key via  cache provider so that each cache items are insulated from each other. 

Please note that http context is unavailable in these cases. The portalID goes for the toss.
Does scheduler, codes in thread will impact this implementation? is there any other way to insulate cache key from mixing up?

Thanks Eric. Yes having seperate instances of is the good choice. This would have seperated the application cache itself. This works. The issue comes with the maintainablity. There is *must have* requirement to run only one instance of code.The application will be hosting hundreds of client with custom modules(hundreds of them) developed for the application. And there is _need_ to run it under one code (including set of dll). The whole idea is to have master set of say X module in one instance of DNN and cater to clients needs by mixing and maching modules. 

Comming back to original Question, Do you see major hurdles in application cache if database are seperate?

You might try using a Differencing Disk set up.  If you created your base install on a parent vhd and then locked it as readonly, you could then proceed with child .vhd from the parent.  Have a look at the html help files of Microsoft Virtual Server, its a free download.

The result should be mutiple unique children environments from a single parent code base, I doubt each child would even be aware of its siblings, much less collide.

I do this to test against multiple access clients,however the scale you are referring to is a whole different ball game.

Let me know how big of mess this makes.

enjoy,

Cliff Nelson